Abstract: The segmentation of the images which have low contrast edges and intensity inhomogeneities is not very accurate or efficient when using level set method of Chan-Vese model（short for CV model）. For intensity inhomogeneity, a bias field will be introduced to revise the local average intensity, also a kernel function will be introduced in the energy functional. In order to improve the efficiency, a Global Convex Segmentation（GCS） model will be conducted based on the above model. Split-Bregman iteration will be applied in the model. The experiment demonstrates that the advanced model is more accurate and efficient.

Key words: Chan-Vese（CV） model, Split-Bregman, Global Convex Segmentation（GCS） model

摘要： 水平集方法中的Chan-Vese模型（简称CV模型）对灰度不均匀及边界对比度低的图像的分割效果不够精确，计算效率也不是很高。针对灰度不均匀引入偏差场来修正CV模型中的区域平均灰度并引入核函数来加权能量泛函。针对计算效率低下的问题，在上述基础上得出其全局凸分割模型（Global Convex Segmentation，GCS），用Split-Bregman迭代求解该模型。实验结果表明：改进后的模型提高了分割精确度和计算效率。

关键词: CV模型, Split-Bregman, 全局凸分割模型（GCS）